moonshotai/Kimi-K2.5 vs kimi-k3
Side-by-side comparison of moonshotai/Kimi-K2.5 and kimi-k3: benchmarks, pricing, context window and capabilities. Both are accessible through Requesty's unified API.

moonshotai/Kimi-K2.5
Input / 1M
$0.45
Output / 1M
$2.25
Context
262K
Model ID
deepinfra/moonshotai/Kimi-K2.5

kimi-k3
Input / 1M
$3.00
Output / 1M
$15.00
Context
1.0M
Model ID
moonshot/kimi-k3
Benchmark comparison
Intelligence Indexreasoning
moonshotai/Kimi-K2.5N/A
kimi-k317.4%
Scores sourced from official model cards, Artificial Analysis, and public leaderboards. Benchmarks measure specific skills and don't capture every aspect of model quality.
Pricing & specifications
| moonshotai/Kimi-K2.5 | kimi-k3 | |
|---|---|---|
| Input price / 1M | $0.45 | $3.00 |
| Output price / 1M | $2.25 | $15.00 |
| Context window | 262K tokens | 1.0M tokens |
| Max output | 131K tokens | 1.0M tokens |
| Vision input | Yes | Yes |
| Tool calling | Yes | Yes |
| Reasoning | Yes | Yes |
| Prompt caching | Yes | Yes |
| Computer use | N/A | N/A |
| Provider | DeepInfra | Moonshot AI |
Questions people ask
Is moonshotai/Kimi-K2.5 better than kimi-k3?
Benchmark data is limited for one or both models. Compare pricing and capabilities in the tables above, and test both on your own workload.
Which is cheaper, moonshotai/Kimi-K2.5 or kimi-k3?
moonshotai/Kimi-K2.5 is cheaper. moonshotai/Kimi-K2.5 costs $0.45/$2.25 per 1M input/output tokens, while kimi-k3 costs $3.00/$15.00.
Can I use moonshotai/Kimi-K2.5 and kimi-k3 through the same API?
Yes. Requesty provides a single OpenAI-compatible API that routes to both. Change just the "model" parameter to switch between "deepinfra/moonshotai/Kimi-K2.5" and "moonshot/kimi-k3", no other code changes needed.
What are the context windows?
moonshotai/Kimi-K2.5 supports up to 262K tokens of context. kimi-k3 supports up to 1.0M tokens. Longer context means you can feed larger documents or codebases in a single prompt, though quality often degrades past 128K for most models.
Switch between moonshotai/Kimi-K2.5 and kimi-k3 with one line of code
Requesty provides a single OpenAI-compatible API for 600+ models. Change the model parameter, not your code.
